Blasting vibration peak velocity prediction model generation and prediction method, device and equipment

By constructing a physical information neural network model combining empirical formulas and using particle swarm optimization algorithm training, a burst vibration peak velocity prediction model is generated, which solves the data dependence problem and achieves efficient prediction effect.

CN120509444APending Publication Date: 2025-08-19WUHAN UNIV OF TECH
View PDF 0 Cites 4 Cited by

Patent Information

Application Number
CN202510478598.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-04-16
Publication Date
2025-08-19

AI Technical Summary

Technical Problem

The blasting vibration peak velocity prediction model generated by the prior art is highly dependent on the number of data and cannot effectively guide the vibration effect control.

Method used

The first physical information neural network model is constructed, the loss function is predicted by combining empirical formulas, and the particle swarm optimization algorithm is optimized and trained to generate the second physical information neural network model, and finally trained to obtain the blast vibration peak velocity prediction model.

Benefits of technology

Good prediction performance can be achieved through less data training, reducing dependence on data, and improving the accuracy and efficiency of the prediction model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120509444A_ABST
    Figure CN120509444A_ABST
Patent Text Reader

Abstract

The invention provides a blast vibration peak velocity prediction model generation method, a blast vibration peak velocity prediction method, a blast vibration peak velocity prediction device and blast vibration peak velocity prediction equipment, and belongs to the technical field of engineering blasting. Loss functions of the first physical information neural network model comprise a physical information neural network prediction loss function and an empirical formula prediction loss function, and the empirical formula prediction loss function is used for calculating a mean square error between a prediction value of a preset empirical formula and a preset training set actual value; training the first physical information neural network model based on a preset training set; optimizing and training the complete first physical information neural network model based on a particle swarm optimization algorithm to obtain a second physical information neural network model; and training the second physical information neural network model based on the training set to obtain a blasting vibration peak velocity prediction model. The technical problem that a prediction model generated in the prior art seriously depends on the number of data is effectively solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of engineering blasting, and in particular to a blasting vibration peak velocity prediction model generation and prediction method, device and equipment. Background Art

[0002] With the continuous development of modern blasting technology, drilling and blasting has been widely used in fields such as mining, subway tunnels, and infrastructure construction. However, blasting operations can cause varying degrees of vibration to buildings near the blasting area. Determining the magnitude of this impact based on factors such as explosive charge and distance has become a major challenge. Traditionally, empirical formulas have been used to predict the peak velocity of blasting vibrations, thereby determining the magnitude of the vibration impact. However, this method has limited prediction accuracy and cannot effectively guide the control of vibration effects. With the development of artificial intelligence technology, an increasing number of machine learning methods are being applied to blasting vibration peak velocity prediction to improve the shortcomings of traditional blasting vibration peak velocity prediction formulas.

[0003] Currently, commonly used machine learning methods for blasting vibration peak velocity include XGBoost, RF, SVR, and BPNN. These methods share a common characteristic: they are all data-driven. They all use monitored data on blasting vibration peak velocity and related influencing factors to train machine learning models and then predict blasting vibration peak velocity. Therefore, the performance of the models is determined to a certain extent by the data obtained. Model performance is highly dependent on the dataset, and obtaining more data for model training requires more manpower and material resources for data monitoring. Summary of the Invention

[0004] In view of this, it is necessary to provide a blasting vibration peak velocity prediction model generation and prediction method, device and equipment to solve the technical problem that the prediction model generated by the existing technology is seriously dependent on the amount of data.

[0005] In order to solve the above technical problems, in a first aspect, the present invention provides a method for generating a blasting vibration peak velocity prediction model, comprising: Constructing a first physical information neural network model, wherein the loss function of the first physical information neural network model includes: a physical information neural network prediction loss function and an empirical formula prediction loss function, wherein the empirical formula prediction loss function is used to calculate the mean square error between the predicted value of the preset empirical formula and the actual value of the preset training set; Training the first physical information neural network model based on a preset training set; Optimizing the first physical information neural network model that has been fully trained based on a particle swarm optimization algorithm to obtain a second physical information neural network model; The second physical information neural network model is trained based on the training set to obtain a final blasting vibration peak velocity prediction model.

[0006] In some embodiments of the present invention, the loss function includes:

[0007] Where, represents the loss function; Represents weight, the value range is [0,1]; Represents the prediction loss function of the physical information neural network model; Represents the empirical formula prediction loss function; The physical information neural network model prediction loss function includes:

[0008] Where, Indicates the number of training samples; Indicates the training samples; represents the actual value of the training sample; Represents the predicted value of the physical information neural network model; The empirical formula prediction loss function includes:

[0009] Where, Indicates the number of training samples; Indicates the training samples; represents the actual value of the training sample; represents the predicted value of the empirical formula; Wherein, the empirical formula includes:

[0010] Where, is the peak velocity of blasting vibration, To monitor the distance, is the resistance line length, is the maximum charge amount for a single hole, is a coefficient related to the terrain and geological conditions between the blasting point and the protected object. It is the attenuation index related to the terrain and geological conditions between the blasting point and the protected object. is the inelastic attenuation factor.

[0011] In some embodiments of the present invention, the training of the first physical information neural network model based on the training set includes: Initializing the weights and biases of the first physical information neural network model; Taking the monitoring distance, the resistance line length and the maximum charge of a single hole as characteristic variables and the blasting vibration peak velocity as the target variable, the training set is input into the initialized first physical information neural network model for training, and the loss function is calculated; The weights and biases of the first physical information neural network model are updated based on the loss function and the Adam optimization algorithm until the value of the loss function reaches a preset loss threshold or the training cycle reaches a preset maximum training cycle value.

[0012] In some embodiments of the present invention, updating the weights and biases of the first physical information neural network model based on the loss function and the Adam optimization algorithm includes: Starting from the output layer of the first physical information neural network model, the gradient of the loss function for each layer in the first physical information neural network model is calculated based on the chain rule, and the gradient is propagated forward layer by layer; The weights and biases of the first physical information neural network model are updated based on the gradient and the Adam optimization algorithm.

[0013] In some embodiments of the present invention, the method of optimizing the first physical information neural network model that has been fully trained based on the particle swarm optimization algorithm to obtain the second physical information neural network model includes: Setting a hyperparameter combination of the first physical information neural network model as particles of a particle swarm optimization algorithm and initializing the particles, wherein the hyperparameter combination includes: a weight of a loss function, a number of neurons in a hidden layer, and a batch size of training data; Training the first physical information neural network model based on the initialized particles and the training set, and calculating a fitness value, wherein the fitness value includes: a mean square error of a particle swarm optimization algorithm; Updating the particles based on a preset speed update formula and a position update formula; Replacing the initialized particles with updated particles, retraining the first physical information neural network model and calculating the fitness value, and iterating until the difference in fitness values between the two trainings is no greater than a preset threshold or the number of iterations reaches a preset maximum number of iterations; The hyperparameter combination represented by the particle with the global minimum fitness in the iterative process is transferred as the optimal hyperparameter combination to the first physical information neural network model for optimization to obtain the second physical information neural network model.

[0014] In some embodiments of the present invention, after obtaining the final blasting vibration peak velocity prediction model, the method further includes: Inputting the preset test set into the blasting vibration peak velocity prediction model to obtain a prediction result; Evaluation indicators of the blasting vibration peak velocity prediction model are calculated based on the test set and the prediction results. The evaluation indicators include: accuracy, mean square error, root mean square error, and determination coefficient.

[0015] In a second aspect, the present invention further provides a method for predicting blasting vibration peak velocity, comprising: Obtaining a blasting vibration peak velocity prediction model based on the blasting vibration peak velocity prediction model generation method described in any one of the above methods; The preset monitoring distance, resistance line length and single-hole maximum charge are input into the blasting vibration peak velocity prediction model to predict the blasting vibration peak velocity.

[0016] In a third aspect, the present invention further provides a device for generating a blasting vibration peak velocity prediction model, comprising: A model construction module is used to construct a first physical information neural network model, wherein the loss function of the first physical information neural network model includes: a physical information neural network prediction loss function and an empirical formula prediction loss function, wherein the empirical formula prediction loss function is used to calculate the mean square error between the predicted value of the preset empirical formula and the actual value of the preset training set; A first training module, configured to train a first physical information neural network model based on a training set; A model optimization module, configured to optimize the first physical information neural network model that has been trained completely based on a particle swarm optimization algorithm to obtain a second physical information neural network model; The second training module is used to train the second physical information neural network model based on the training set to obtain a final blasting vibration peak velocity prediction model.

[0017] In some embodiments of the present invention, the apparatus further comprises: The model testing module is used to input a preset test set into the blasting vibration peak velocity prediction model to obtain a prediction result, and calculate the evaluation index of the blasting vibration peak velocity prediction model based on the test set and the prediction result. The evaluation index includes: accuracy, mean square error, root mean square error and determination coefficient.

[0018] In a fourth aspect, the present invention further provides a device comprising: Memory, used to store programs; A processor is coupled to the memory and is used to execute the program stored in the memory to implement the steps of the blasting vibration peak velocity prediction model generation method or the blasting vibration peak velocity prediction method described in any one of the above method items.

[0019] The beneficial effects of the present invention are as follows: the present invention provides a method for generating a blasting vibration peak velocity prediction model, which first combines an empirical formula as physical prior knowledge with a loss function in a neural network to construct a first physical information neural network model. The first physical information neural network model is then optimized based on a particle swarm optimization algorithm during training, and a second physical information neural network model with better performance is obtained based on the architecture of the first physical information neural network model. Finally, the second physical information neural network model is trained again to obtain a final blasting vibration peak prediction model. By incorporating the empirical formula into the loss function of the neural network to constrain the training process of the neural network, the present invention can achieve good prediction performance even after training with less data, thereby eliminating the need for large amounts of data to train the model, effectively solving the technical problem that prediction models generated by existing technologies are heavily dependent on the amount of data. BRIEF DESCRIPTION OF THE DRAWINGS In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0020] Figure 1 A schematic flow chart of an embodiment of a method for generating a blasting vibration peak velocity prediction model provided by the present invention; Figure 2 for Figure 1 A flow chart of an embodiment of step S102; Figure 3 for Figure 2 A flow chart of an embodiment of step S1023; Figure 4 for Figure 1 A flow chart of an embodiment of step S103; Figure 5 A schematic flow chart of an embodiment of a method for predicting blasting vibration peak velocity provided by the present invention; Figure 6 A schematic diagram of an embodiment of a device for generating a blasting vibration peak velocity prediction model provided by the present invention; Figure 7 This is a structural diagram of an embodiment of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0022] In the description of the embodiments of the present invention, unless otherwise specified, "plurality" means two or more. "And / or" describes the association relationship between associated objects, indicating that three relationships can exist. For example, "A and / or B" can mean: A exists alone, A and B exist simultaneously, or B exists alone.

[0023] The terms "first," "second," and so on, used in the embodiments of the present invention are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, technical features designated as "first" or "second" may explicitly or implicitly include at least one such feature.

[0024] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present invention. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute a separate or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0025] The present invention provides a blasting vibration peak velocity prediction model generation and prediction method, device and equipment, which are described below respectively.

[0026] Before building the model, we first acquired accurate blasting vibration test data. Specifically, we used a blasting vibration tester to monitor the blasting vibration signals at different monitoring distances. The three-dimensional blasting vibration velocities obtained were then synthesized. The maximum value within the synthesized signal was then identified to determine the peak blasting vibration velocity at each monitoring distance. Furthermore, we recorded the corresponding monitoring distance, resistance line length, and maximum charge per hole. This testing process was repeated multiple times to build a data set.

[0027] The characteristic variables of the model are determined to be the monitoring distance, the length of the resistance line, and the maximum charge per hole, and the target variable is the peak velocity of the blasting vibration. The characteristic variable data is standardized so that different characteristic variables have the same scale and eliminate the influence of dimension. The standardization formula is as follows: (1) In formula (1), is the standardized data; is the original data; is the mean of the data; is the variance of the data.

[0028] Then, the obtained data is randomly divided, 80% of which is the training set for training the model, and 20% is the test set for testing the performance of the model.

[0029] Figure 1 A flow chart of an embodiment of the method for generating a blasting vibration peak velocity prediction model provided by the present invention is shown as follows: Figure 1 As shown in FIG, the method for generating a blasting vibration peak velocity prediction model includes: S101. Construct a first physical information neural network model.

[0030] First, set the structure of the neural network and determine the number of hidden layers M , the number of neurons in the input layer is N_input , whose value depends on the number of feature variables, and the number of neurons in the output layer is 1. Set the activation function of the hidden layer. Common activation functions include ReLU, sigmoid, tanh, etc. The output layer does not use an activation function. In addition, determine the learning rate, regularization parameter, and number of iterations. Set the maximum number of iterations of the particles, the number of populations, inertia weights, learning factors, hyperparameter search range, and randomly initialize a group of particles. Determine the weights in the neural network loss function based on the initialized particles. w , the number of neurons in the hidden layer N_hidden , the batch size of the training data batch_size .

[0031] It should be noted that the loss function of the first physical information neural network model includes: a physical information neural network prediction loss function and an empirical formula prediction loss function, wherein the empirical formula prediction loss function is used to calculate the mean square error between the predicted value of the preset empirical formula and the actual value of the preset training set.

[0032] Specifically, the loss function is: (2) In formula (2), represents the loss function; Represents weight, the value range is [0,1]; Represents the prediction loss function of the physical information neural network model; Represents the empirical formula prediction loss function; Among them, the prediction loss function of the physical information neural network model includes: (3) In formula (3), Indicates the number of training samples; Indicates the training samples; represents the actual value of the training sample; Represents the predicted value of the physical information neural network model; The empirical formula prediction loss function includes: (4) In formula (4), Indicates the number of training samples; Indicates the training samples; represents the actual value of the training sample; represents the predicted value of the empirical formula; Wherein, the empirical formula includes: (5) In formula (5), is the peak velocity of blasting vibration, To monitor the distance, is the resistance line length, is the maximum charge amount for a single hole, is a coefficient related to the terrain and geological conditions between the blasting point and the protected object. It is the attenuation index related to the terrain and geological conditions between the blasting point and the protected object. is the inelastic attenuation factor.

[0033] S102: Train the first physical information neural network model based on the training set.

[0034] It should be noted that the training set includes test data from several rounds of blasting vibration peak velocity tests.

[0035] Furthermore, the test data includes: peak blasting vibration velocity, as well as monitoring distance, resistance line length and maximum charge per single hole corresponding to the blasting vibration velocity.

[0036] like Figure 2 In some embodiments of the present invention, step S102 includes: S1021. Initialize the weights and biases of the first physical information neural network model.

[0037] S1022. Using the monitoring distance, resistance line length, and maximum charge per hole as characteristic variables and the blasting vibration peak velocity as the target variable, the training set is input into the initialized first physical information neural network model for training, and the loss function is calculated.

[0038] S1023. Update the weights and biases of the first physical information neural network model based on the loss function and the Adam optimization algorithm until the value of the loss function reaches a preset loss threshold or the training cycle reaches a preset maximum training cycle value.

[0039] Furthermore, if Figure 3 , step S1023 includes: S10231. Starting from the output layer of the first physical information neural network model, calculate the gradient of the loss function for each layer in the first physical information neural network model based on the chain rule, and propagate the gradient forward layer by layer; S10232. Update the weights and biases of the first physical information neural network model based on the gradient and Adam optimization algorithm.

[0040] In a specific embodiment of the present invention, the weights and biases of the physical information neural network are first initialized randomly, and then the input data is passed from the input layer to the output layer. Each neuron in each layer receives input from the previous layer, calculates the output using an activation function, and passes the result to the next layer. During calculation, the output of the neuron is the weighted sum of the input value and the weight, plus the bias term, and then converted by the activation function. The formula is as follows: (6) In formula (6), x i For the i Input data, w i For the i The weights of the input features, b is the bias term, F is the activation function.

[0041] Secondly, the difference between the output value of the physical information neural network and the actual value is calculated, that is, the loss function value. The loss function of the physical information neural network consists of two parts. One part is the prediction loss of the physical information neural network model. LOSS MSE , part of which is the empirical formula for predicting losses LOSS Formula .

[0042] Then, using the chain rule, the gradient of the loss function for each layer is calculated, starting from the output layer and propagating forward layer by layer. After calculating the gradient, the Adam optimization algorithm is used to update the weights and biases in the network, so that the value of the loss function gradually decreases, thereby improving the network's prediction accuracy. The formula for updating parameters in the Adam optimization algorithm is as follows: (7) In formula (7), It ist Parameters at the iteration; It is t +1 updated parameters at iteration 1; η is the learning rate, which controls the size of each update step; is the bias-corrected version of the first moment of the gradient, representing the cumulative mean of the gradient; is the bias-corrected version of the second moment of the gradient, representing the cumulative squared mean of the gradient; ε is a small constant.

[0043] The deviation correction formulas for the first-order moment and the second-order moment are as follows: (8) (9) In formulas (8) and (9), m t is the current estimate of the first moment, is the decay rate of the first-order moment; v t is the current estimate of the second moment, is the decay rate of the second-order moment, and t is the current number of iterations.

[0044] The calculation formulas for the first-order moment and second-order moment of the gradient are as follows:

[0045] In formulas (10) and (11), g t is the current gradient, 、 are the hyperparameters that control the decay rate.

[0046] Finally, the training process is repeated for multiple epochs until the loss function reaches a smaller value or the maximum number of epochs is reached. Each epoch consists of a complete pass through the training data, and in each epoch, the network weights are updated based on the gradient and loss function.

[0047] S103. Optimize the first physical information neural network model that has been trained completely based on a particle swarm optimization algorithm to obtain a second physical information neural network model.

[0048] like Figure 4 In some embodiments of the present invention, step S103 includes: S1031. Set the hyperparameter combination of the first physical information neural network model as particles of the particle swarm optimization algorithm, and initialize the particles.

[0049] Among them, the hyperparameter combination includes: the weight of the loss function, the number of neurons in the hidden layer, and the batch size of the training data.

[0050] S1032: Train the first physical information neural network model based on the initialized particles and the training set, and calculate the fitness value.

[0051] Preferably, the fitness value includes: the mean square error of the particle swarm optimization algorithm.

[0052] S1033. Update the particles based on the preset speed update formula and position update formula.

[0053] Specifically, the velocity and coordinates of the particle are iteratively updated according to the velocity update formula and the position update formula, and the hyperparameter combination is updated. The velocity update formula and the position update formula are as follows: (12) (13) In formulas (12) and (13), v ij For the i The particle in j speed in one dimension; x ij For the i The particle in j Position in one dimension; p ij For the i The local optimal position of each particle; p gj is the global optimal position of the particle swarm; w is the inertia weight; c 1 and c 2 are learning factors; r 1 and r 2 is a random number in the range [0,1].

[0054] S1034. Replace the initialized particles with the updated particles, retrain the first physical information neural network model and calculate the fitness value, and iterate until the difference in fitness values between the two previous and subsequent trainings is no greater than a preset threshold or the number of iterations reaches a preset maximum number of iterations.

[0055] S1035. The hyperparameter combination represented by the particle with the global minimum fitness in the iterative process is transferred as the optimal hyperparameter combination to the first physical information neural network model for optimization to obtain a second physical information neural network model.

[0056] It should be noted that in this embodiment, based on the fitness value of each particle after moving to the new position, its individual and group historical optimal positions are dynamically adjusted according to the formula. When the particle's fitness tends to stabilize or the number of iterations reaches the maximum, the iteration cycle is terminated to obtain the optimal parameter combination.

[0057] The formula for updating the individual's historical best position is as follows: (14) The formula for updating the historical optimal position of the group is as follows: (15) In formulas (14) and (15), It is i The particle in t The historical best position of the generation, is the current particle position, f is the objective function, The group is t +1 generation's historical best position.

[0058] S104: Training the second physical information neural network model based on the training set to obtain a final blasting vibration peak velocity prediction model.

[0059] After the blasting vibration peak velocity prediction model is trained, the model performance needs to be tested. By observing the evaluation indicators during the model testing process, the performance of the model can be tested to see whether it is good. In some embodiments of the present invention, after step S104, the following steps are further included: S105 , inputting the preset test set into the blasting vibration peak velocity prediction model to obtain a prediction result.

[0060] S107. Calculate the evaluation index of the blasting vibration peak velocity prediction model based on the test set and the prediction results.

[0061] Preferably, the evaluation indicators include: accuracy, mean square error, root mean square error and determination coefficient.

[0062] Specifically, the evaluation index of the model is calculated by the actual value and the predicted value. The model evaluation index includes accuracy (AR), mean square error (MSE), root mean square error (RMSE), coefficient of determination (R 2 ). The specific calculation formula is as follows: (16) (17) (18) (19) In formulas (16)-(19), i It is i test samples, y i is the actual value, is the predicted value, is the mean of the predicted values, n is the number of all test samples.

[0063] Compared to existing technologies, the present invention provides a method for generating a blasting vibration peak velocity prediction model. This method first combines an empirical formula as physical prior knowledge with a loss function in a neural network to construct a first physical information neural network model. During training, the first physical information neural network model is then optimized using a particle swarm optimization algorithm. Building on the architecture of the first physical information neural network model, a second physical information neural network model with improved performance is obtained. Finally, the second physical information neural network model is trained again to obtain the final blasting vibration peak velocity prediction model. By incorporating the empirical formula into the neural network's loss function to constrain the neural network training process, the present invention enables the model to achieve good prediction performance even after training with relatively little data, eliminating the need for extensive data training. This effectively addresses the technical issue of prediction models generated by existing technologies being heavily dependent on data volume. Furthermore, after obtaining a fully trained model, the present invention also performs real-time model evaluation using a test set and outputs evaluation metrics, allowing users to observe the model's performance in real time and make timely adjustments, further improving the accuracy of the blasting vibration peak velocity prediction model.

[0064] like Figure 5 In a second aspect, the present invention further provides a method for predicting blasting vibration peak velocity, comprising: S501, obtaining a blasting vibration peak velocity prediction model based on the blasting vibration peak velocity prediction model generation method of any one of the above methods; S502: Input the preset monitoring distance, resistance line length and single-hole maximum charge into the blasting vibration peak velocity prediction model to predict the blasting vibration peak velocity.

[0065] It is understood that the method for generating a blasting vibration peak velocity prediction model provided by the present invention constrains the neural network training process by incorporating an empirical formula into the neural network's loss function. This allows the model to achieve good prediction performance even after training with relatively little data, eliminating the need for extensive data training. Furthermore, after obtaining a fully trained model, the present invention uses a test set to evaluate the model in real time and outputs the resulting evaluation metrics, allowing users to observe the model's performance in real time and make timely adjustments to the model, further improving the accuracy of the blasting vibration peak velocity prediction model. Therefore, in this embodiment, the blasting vibration peak velocity prediction model obtained using the blasting vibration peak velocity prediction model described above can achieve more accurate predictions using relatively little test data, demonstrating good generalizability.

[0066] like Figure 6 In a third aspect, the present invention further provides a blasting vibration peak velocity prediction model generation device 60, comprising: A model construction module 610 is configured to construct a first physical information neural network model. The loss function of the first physical information neural network model includes a physical information neural network prediction loss function and an empirical formula prediction loss function. The empirical formula prediction loss function is configured to calculate the mean square error between the predicted value of a preset empirical formula and the actual value of a preset training set. A first training module 620 is used to train a first physical information neural network model based on a training set; A model optimization module 630 is configured to optimize the first physical information neural network model that has been trained completely based on a particle swarm optimization algorithm to obtain a second physical information neural network model; A second training module 640 is used to train the second physical information neural network model based on the training set to obtain a final blasting vibration peak velocity prediction model; The model testing module 650 is used to input a preset test set into the blasting vibration peak velocity prediction model to obtain a prediction result, and calculate the evaluation indicators of the blasting vibration peak velocity prediction model based on the test set and the prediction result. The evaluation indicators include: accuracy, mean square error, root mean square error and determination coefficient.

[0067] like Figure 7 In a fourth aspect, the present invention further provides an electronic device 70, comprising: Memory 710, used for storing programs; The processor 720 is coupled to the memory 710 and is configured to execute the program stored in the memory 710 to implement the steps of the blasting vibration peak velocity prediction model generation method or the blasting vibration peak velocity prediction method described in any one of the above methods.

[0068] It should be understood that, when the processor 720 executes the blasting vibration peak velocity prediction model generation program or the blasting vibration peak velocity prediction program in the memory 710 , in addition to the above functions, it can also implement other functions. For details, please refer to the description of the corresponding method embodiment above.

[0069] Furthermore, the embodiment of the present invention does not specifically limit the type of the electronic device 70 mentioned. The electronic device 70 may be a portable electronic device such as a mobile phone, a tablet computer, a personal digital assistant (PDA), a wearable device, or a laptop computer. Exemplary embodiments of portable electronic devices include but are not limited to portable electronic devices equipped with iOS, Android, Microsoft, or other operating systems. The above-mentioned portable electronic devices may also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of the present invention, the electronic device 70 may not be a portable electronic device, but a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0070] Those skilled in the art will appreciate that all or part of the process flow of the above-described method embodiment can be implemented by instructing related hardware (such as a processor, controller, etc.) through a computer program, and the computer program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a magnetic disk, an optical disk, a read-only memory, or a random access memory.

[0071] The above is a detailed introduction to the generation and prediction method, device and equipment of a blasting vibration peak velocity prediction model provided by the present invention. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method and core ideas of the present invention. At the same time, for those skilled in the art, according to the ideas of the present invention, there may be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for generating a blasting vibration peak velocity prediction model, characterized in that: include: Constructing a first physical information neural network model, wherein the loss function of the first physical information neural network model includes: a physical information neural network prediction loss function and an empirical formula prediction loss function, wherein the empirical formula prediction loss function is used to calculate the mean square error between the predicted value of the preset empirical formula and the actual value of the preset training set; Training a first physical information neural network model based on the training set; Optimizing the first physical information neural network model that has been fully trained based on a particle swarm optimization algorithm to obtain a second physical information neural network model; The second physical information neural network model is trained based on the training set to obtain a final blasting vibration peak velocity prediction model.

2. The method for generating a blasting vibration peak velocity prediction model according to claim 1, wherein: The loss function includes: Where, represents the loss function; Represents weight, the value range is [0,1]; Represents the prediction loss function of the physical information neural network model; Represents the empirical formula prediction loss function; The physical information neural network model prediction loss function includes: Where, Indicates the number of training samples; Indicates the training samples; represents the actual value of the training sample; Represents the predicted value of the physical information neural network model; The empirical formula prediction loss function includes: Where, Indicates the number of training samples; Indicates the training samples; represents the actual value of the training sample; represents the predicted value of the empirical formula; Wherein, the empirical formula includes: Where, is the peak velocity of blasting vibration, To monitor the distance, is the resistance line length, is the maximum charge amount for a single hole, is a coefficient related to the terrain and geological conditions between the blasting point and the protected object. It is the attenuation index related to the terrain and geological conditions between the blasting point and the protected object. is the inelastic attenuation factor.

3. The method for generating a blasting vibration peak velocity prediction model according to claim 1, wherein: The training of the first physical information neural network model based on the training set includes: Initializing the weights and biases of the first physical information neural network model; Taking the monitoring distance, the resistance line length and the maximum charge of a single hole as characteristic variables and the blasting vibration peak velocity as the target variable, the training set is input into the initialized first physical information neural network model for training, and the loss function is calculated; The weights and biases of the first physical information neural network model are updated based on the loss function and the Adam optimization algorithm until the value of the loss function reaches a preset loss threshold or the training cycle reaches a preset maximum training cycle value.

4. The method for generating a blasting vibration peak velocity prediction model according to claim 3, wherein: The updating of the weights and biases of the first physical information neural network model based on the loss function and the Adam optimization algorithm includes: Starting from the output layer of the first physical information neural network model, the gradient of the loss function for each layer in the first physical information neural network model is calculated based on the chain rule, and the gradient is propagated forward layer by layer; The weights and biases of the first physical information neural network model are updated based on the gradient and the Adam optimization algorithm.

5. The method for generating a blasting vibration peak velocity prediction model according to claim 1, wherein: The method of optimizing and training the first physical information neural network model based on the particle swarm optimization algorithm to obtain the second physical information neural network model includes: Setting a hyperparameter combination of the first physical information neural network model as particles of a particle swarm optimization algorithm and initializing the particles, wherein the hyperparameter combination includes: a weight of a loss function, a number of neurons in a hidden layer, and a batch size of training data; Training the first physical information neural network model based on the initialized particles and the training set, and calculating a fitness value, wherein the fitness value includes: a mean square error of a particle swarm optimization algorithm; Updating the particles based on a preset speed update formula and a position update formula; Replacing the initialized particles with updated particles, retraining the first physical information neural network model and calculating the fitness value, and iterating until the difference in fitness values between the two trainings is no greater than a preset threshold or the number of iterations reaches a preset maximum number of iterations; The hyperparameter combination represented by the particle with the global minimum fitness in the iterative process is transferred as the optimal hyperparameter combination to the first physical information neural network model for optimization to obtain the second physical information neural network model.

6. The method for generating a blasting vibration peak velocity prediction model according to claim 1, wherein: After obtaining the final blasting vibration peak velocity prediction model, the method further includes: Inputting a preset test set into the blasting vibration peak velocity prediction model to obtain a prediction result, wherein the test set includes: test data of several rounds of blasting vibration peak velocity tests; Evaluation indicators of the blasting vibration peak velocity prediction model are calculated based on the test set and the prediction results. The evaluation indicators include: accuracy, mean square error, root mean square error, and determination coefficient.

7. A method for predicting blasting vibration peak velocity, characterized in that: include: Obtaining a blasting vibration peak velocity prediction model based on the blasting vibration peak velocity prediction model generation method according to any one of claims 1 to 6; The preset monitoring distance, resistance line length and single-hole maximum charge are input into the blasting vibration peak velocity prediction model to predict the blasting vibration peak velocity.

8. A device for generating a blasting vibration peak velocity prediction model, characterized in that: include: A model construction module is used to construct a first physical information neural network model, wherein the loss function of the first physical information neural network model includes: a physical information neural network prediction loss function and an empirical formula prediction loss function, wherein the empirical formula prediction loss function is used to calculate the mean square error between the predicted value of the preset empirical formula and the actual value of the preset training set; A first training module, configured to train a first physical information neural network model based on a training set; A model optimization module, configured to optimize the first physical information neural network model that has been trained completely based on a particle swarm optimization algorithm to obtain a second physical information neural network model; The second training module is used to train the second physical information neural network model based on the training set to obtain a final blasting vibration peak velocity prediction model.

9. The blasting vibration peak velocity prediction model generation device according to claim 8, characterized in that: The device further comprises: The model testing module is used to input a preset test set into the blasting vibration peak velocity prediction model to obtain a prediction result, and calculate the evaluation index of the blasting vibration peak velocity prediction model based on the test set and the prediction result. The evaluation index includes: accuracy, mean square error, root mean square error and determination coefficient.

10. A device, characterized in that include: Memory, used to store programs; A processor, coupled to the memory, is configured to execute the program stored in the memory to implement the steps of the blasting vibration peak velocity prediction model generation method described in any one of claims 1 to 6 or the blasting vibration peak velocity prediction method described in claim 7.

Citation Information

Cited By

  • Shock absorber valve system adjusting method based on artificial intelligence technology, product, equipment and storage medium

    CN120800838A

  • A shock absorber valve system tuning method, product, device and storage medium based on artificial intelligence technology

    CN120800838B

  • Fan blade vibration response prediction method and system

    CN121071626A

  • A method and system for predicting the vibration response of wind turbine blades

    CN121071626B